EP3663988A1 - Artificial neuron for neuromorphic chip with resistive synapses - Google Patents
Artificial neuron for neuromorphic chip with resistive synapses Download PDFInfo
- Publication number
- EP3663988A1 EP3663988A1 EP19212046.7A EP19212046A EP3663988A1 EP 3663988 A1 EP3663988 A1 EP 3663988A1 EP 19212046 A EP19212046 A EP 19212046A EP 3663988 A1 EP3663988 A1 EP 3663988A1
- Authority
- EP
- European Patent Office
- Prior art keywords
- circuit
- synapse
- synaptic
- reading
- voltage
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 210000000225 synapse Anatomy 0.000 title claims abstract description 114
- 210000002569 neuron Anatomy 0.000 title claims abstract description 74
- 230000000946 synaptic effect Effects 0.000 claims abstract description 91
- 230000010354 integration Effects 0.000 claims abstract description 47
- 239000012528 membrane Substances 0.000 claims abstract description 40
- 230000015654 memory Effects 0.000 claims abstract description 29
- 230000001242 postsynaptic effect Effects 0.000 claims abstract description 13
- 230000002401 inhibitory effect Effects 0.000 claims description 19
- 230000002964 excitative effect Effects 0.000 claims description 16
- 230000003518 presynaptic effect Effects 0.000 claims description 13
- 230000001747 exhibiting effect Effects 0.000 abstract 1
- 210000004027 cell Anatomy 0.000 description 15
- 230000004913 activation Effects 0.000 description 8
- 230000000694 effects Effects 0.000 description 8
- 101100368144 Mus musculus Synb gene Proteins 0.000 description 7
- 230000010287 polarization Effects 0.000 description 7
- 230000003068 static effect Effects 0.000 description 7
- 210000005215 presynaptic neuron Anatomy 0.000 description 6
- 210000001787 dendrite Anatomy 0.000 description 4
- 230000005284 excitation Effects 0.000 description 4
- 230000008520 organization Effects 0.000 description 4
- 230000003071 parasitic effect Effects 0.000 description 4
- 230000002441 reversible effect Effects 0.000 description 4
- 230000000638 stimulation Effects 0.000 description 4
- 208000031968 Cadaver Diseases 0.000 description 3
- 238000013528 artificial neural network Methods 0.000 description 3
- 210000003050 axon Anatomy 0.000 description 3
- 230000008901 benefit Effects 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 230000005764 inhibitory process Effects 0.000 description 3
- 230000002829 reductive effect Effects 0.000 description 3
- 230000036982 action potential Effects 0.000 description 2
- 230000003213 activating effect Effects 0.000 description 2
- 239000003990 capacitor Substances 0.000 description 2
- 210000000170 cell membrane Anatomy 0.000 description 2
- 239000004020 conductor Substances 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000002347 injection Methods 0.000 description 2
- 239000007924 injection Substances 0.000 description 2
- 150000002500 ions Chemical class 0.000 description 2
- 238000000034 method Methods 0.000 description 2
- 210000004205 output neuron Anatomy 0.000 description 2
- 238000011176 pooling Methods 0.000 description 2
- 230000001902 propagating effect Effects 0.000 description 2
- 239000000243 solution Substances 0.000 description 2
- 241001080024 Telles Species 0.000 description 1
- 230000009471 action Effects 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 230000002457 bidirectional effect Effects 0.000 description 1
- 238000013529 biological neural network Methods 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 230000015572 biosynthetic process Effects 0.000 description 1
- 210000005056 cell body Anatomy 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 230000008878 coupling Effects 0.000 description 1
- 238000010168 coupling process Methods 0.000 description 1
- 238000005859 coupling reaction Methods 0.000 description 1
- 230000003247 decreasing effect Effects 0.000 description 1
- 230000002999 depolarising effect Effects 0.000 description 1
- 230000001627 detrimental effect Effects 0.000 description 1
- 235000021183 entrée Nutrition 0.000 description 1
- 230000002349 favourable effect Effects 0.000 description 1
- 238000007667 floating Methods 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 210000002364 input neuron Anatomy 0.000 description 1
- 239000012212 insulator Substances 0.000 description 1
- 230000000670 limiting effect Effects 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 239000002244 precipitate Substances 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 230000002035 prolonged effect Effects 0.000 description 1
- 230000000644 propagated effect Effects 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 230000000284 resting effect Effects 0.000 description 1
- 238000009738 saturating Methods 0.000 description 1
- 230000001960 triggered effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/049—Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/06—Physical realisation, i.e. hardware implementation of neural networks, neurons or parts of neurons
- G06N3/063—Physical realisation, i.e. hardware implementation of neural networks, neurons or parts of neurons using electronic means
- G06N3/065—Analogue means
-
- G—PHYSICS
- G11—INFORMATION STORAGE
- G11C—STATIC STORES
- G11C11/00—Digital stores characterised by the use of particular electric or magnetic storage elements; Storage elements therefor
- G11C11/21—Digital stores characterised by the use of particular electric or magnetic storage elements; Storage elements therefor using electric elements
- G11C11/34—Digital stores characterised by the use of particular electric or magnetic storage elements; Storage elements therefor using electric elements using semiconductor devices
-
- G—PHYSICS
- G11—INFORMATION STORAGE
- G11C—STATIC STORES
- G11C11/00—Digital stores characterised by the use of particular electric or magnetic storage elements; Storage elements therefor
- G11C11/54—Digital stores characterised by the use of particular electric or magnetic storage elements; Storage elements therefor using elements simulating biological cells, e.g. neuron
-
- G—PHYSICS
- G11—INFORMATION STORAGE
- G11C—STATIC STORES
- G11C13/00—Digital stores characterised by the use of storage elements not covered by groups G11C11/00, G11C23/00, or G11C25/00
- G11C13/0002—Digital stores characterised by the use of storage elements not covered by groups G11C11/00, G11C23/00, or G11C25/00 using resistive RAM [RRAM] elements
-
- G—PHYSICS
- G11—INFORMATION STORAGE
- G11C—STATIC STORES
- G11C13/00—Digital stores characterised by the use of storage elements not covered by groups G11C11/00, G11C23/00, or G11C25/00
- G11C13/0002—Digital stores characterised by the use of storage elements not covered by groups G11C11/00, G11C23/00, or G11C25/00 using resistive RAM [RRAM] elements
- G11C13/0021—Auxiliary circuits
- G11C13/004—Reading or sensing circuits or methods
Definitions
- the field of the invention is that of neuromorphic chips with artificial neural networks exploiting synapses with resistive memory.
- ions pass through the cell membrane.
- the imbalance of charges between the inside and the outside of the cell induces a voltage difference on either side of the membrane.
- This membrane tension exceeds a certain level, that is to say when the cell is sufficiently excited, the neuron experiences a brutal exchange of ions.
- This variation called action potential (“action potential” or “spike” in English), propagates along the axon, towards the synaptic buttons which constitute the outputs of the neuron. Seen from outside the cell, these "spikes” constitute the electrical activity of the neuron.
- each neuron is connected to several thousand others via as many synapses.
- the term synapse designates the connection between the axon termination of a so-called presynaptic neuron and a dendrite of a so-called postsynaptic neuron.
- the influence of the presynaptic neuron on the postsynaptic neuron is balanced by the weight of the synapse which can be excitatory or inhibitory.
- a presynaptic spike charges the membrane tension of the postsynaptic neuron and precipitates the generation of a postsynaptic action potential.
- a Presynaptic "spike” has the effect of depolarizing the postsynaptic membrane and delaying the onset of a postsynaptic action potential.
- Artificial neural networks are used in different fields of signal processing, such as data classification, image recognition or decision-making. They are inspired by the biological neural networks of which they imitate the functioning and are essentially composed of artificial neurons interconnected by synapses which can be implemented by resistive components whose conductance varies according to the voltage applied to their terminals.
- resistive components memristors or RRAM are ideal candidates for the implementation of synapses.
- the variable resistance of these components can be increased (so-called Reset operation) or decreased ( Set operation) if relatively high electrical quantities (voltage and / or current) are applied. If you simply want to read the value of their resistance without modifying it ( Read operation), you must apply relatively small electrical quantities.
- the integration of resistive synapses often takes the form of a memory plane, which is called “synaptic plane", in which the synapses are arranged in a network with transverse lines and columns which makes it possible to connect a layer of input neurons. (presynaptic neurons) and a layer of output neurons (postsynaptic neurons).
- Each synapse has a synapse activation terminal and a propagation terminal for a synaptic signal.
- the terminals for activating synapses on the same line are linked together via a word line (or “Word-Line”), and the terminals for propagating synapses in the same column are connected between and connected to an artificial neuron via a bit line (or "Bit-Line").
- a word line is used to inject a voltage pulse into the synapses of the corresponding line and the bit lines are the outputs of these synapses.
- each bit line propagates a current weighted by the value of the corresponding resistive memory towards a postynaptic neuron.
- a resistive memory synapse generally takes the form of a 1T1R cell composed of a variable resistor M and an access transistor T used to regulate the writing currents and whose grid forms the synapse activation terminal.
- 1T1R cells comprising three word lines WL1-WL3, eight bit lines BL1-BL8 each intended to be connected to an output neuron and a source line (or “Source-Line”) SL connected to all synapses.
- Source-Line source line
- each 1S1R cell is a dipole, the terminals of which constitute the activation and propagation terminals, and the synaptic plane is organized as shown in figure 2 where we took the example of a plan comprising three word lines WL1-WL3 and four bit lines BL1-BL4, and where each synapse is taken between a voltage on the corresponding word line and a voltage on its corresponding bit line.
- this organization offers an “all-to-all” connection according to which all the presynaptic neurons are connected via a synapse to all the postsynaptic neurons.
- this “all-to-all” connection can lead to a mainly zero synaptic matrix and therefore to low energy efficiency.
- This organization also requires having to process zero weights at the risk of having a network composed mainly of parasitic synapses, making the analog integration of synaptic weights even more complex.
- RRAM memory which is composed of an insulator wedged between two layers of conductors.
- OxRAM memory which is composed of an insulator wedged between two layers of conductors.
- a conductive wire grows from one of the conductors towards the other, thus reducing the total resistance of the dipole.
- a Set operation lengthens the wire and reduces the resistance, while a Reset operation has the opposite effect.
- Reading an OxRAM memory presents a difficulty, that of parasitic writing by gradually changing the value of the resistance over successive readings. To limit this risk, it is advisable to apply a maximum reading voltage of 100 mV.
- V spike_pre limits the membrane voltage V mem , because the membrane capacity C mem is charged through the resistor R synb on which V spike_pre is applied. In other words, the membrane tension is worth at most the tension V spike_pre .
- V mem being an analog calculation element therefore subject in particular to problems of noise, technological variations, capacitive coupling, its voltage range cannot be reduced to a value as low as 100 mV without greatly increasing the sources of errors. Calculation.
- RRAMs devices often have a low resistance range (with low values of the order of k ⁇ ) and that we can therefore be confronted with time constants of the cut R syn - C mem of l nanosecond order or less.
- the reading impulse must also be propagated over the entire width of the synaptic plane. The shorter this impulse, the less “clean” it is in the end.
- such a pulse is difficult to propagate in active analog circuits responsible for integrating the read current without being confronted with problems of bandwidth and / or distortion.
- the invention aims to provide a solution to meet one and / or the other of the above constraints.
- an artificial neuron for a neuromorphic chip comprising a synapse with resistive memory representative of a synaptic weight, the artificial neuron comprising an integration circuit which comprises an accumulator of synaptic weights at the terminals of which a membrane tension is established. and a comparator configured to emit a postsynaptic pulse in the event of a threshold being crossed by the membrane voltage.
- the invention relates to an artificial neuron NA for neuromorphic chip comprising a synapse with resistive memory representative of a synaptic weight.
- the synapse has a terminal for activating the synapse and a terminal for propagating a synaptic signal.
- the propagation terminal propagates a synaptic signal representative of the value of the resistive memory (ie the synaptic weight) towards the artificial neuron NA via a BL bit line.
- the artificial neuron includes an integration circuit CI which includes an accumulator of synaptic weights ACC at the terminals of which a membrane voltage V mem is established and a comparator COMP configured to emit a postsynaptic pulse SO in the event of crossing a threshold V threshold by the membrane voltage V mem .
- the invention proposes to decouple the reading of the synaptic weight from its integration while avoiding digitizing the reading of the synaptic weight.
- the different constraints can thus be distributed according to whether they are inherited from resistive memory technology or from the functioning of the analog neuron.
- the invention proposes more particularly that the artificial neuron NA furthermore comprises a reading circuit CL configured to impose on the synapse a reading voltage independent of the membrane voltage V mem and to supply to the integration circuit CI an analog quantity representative of the synaptic weight PSa.
- the reading circuit CL thus makes it possible to impose on the synapse a reading voltage having an identical polarization for each reading regardless of the state of the membrane voltage V mem .
- This reading circuit also makes it possible to impose a sufficiently low reading voltage on the synapse to avoid any risk of parasitic writing.
- the result of the reading namely the analog quantity representative of the synaptic weight PSa, is transmitted to the integration circuit.
- this quantity can take different forms, such as a voltage, a current, a charge or a duration for example.
- the integration circuit Cl thus decoupled from the reading, can thus propose a wide range of voltage for the membrane voltage, independent of the specifications of the synapse.
- synapses which can be both excitatory and inhibitory, and which must therefore be capable of both injecting and drawing current from the neuron. Since the synapses are resistive here and the current they control depends on their resistance, two levels of synapse supply voltage are provided, a positive and a negative.
- the figure 6 provides an example of such an excitatory and inhibitory synapse, called synapse 2R1C.
- the synapse comprises an excitatory component CE and an inhibitory component CI arranged in series, with the propagation terminal Bp of the synapse as the midpoint, between a positive supply voltage V read and a negative supply voltage -V read .
- the capacity C load is part of the reading circuit of a neuron according to a possible embodiment of the invention and is therefore shared by all the synapses of the same column.
- the weight of the synapse is encoded as follows. If we want an exciting synapse, the resistance R inhi is adjusted to its maximum resistance level, while the resistance R exci ( ⁇ R inhi ) modulates the positive weight of the synapse. The lower the resistance R exci , the greater the synaptic weight, since this results in a higher synaptic current.
- the resistance R exci is adjusted to its maximum resistance level, while the resistance R inhi ( ⁇ R exci ) modulates the negative weight of the synapse
- the diagram composed of the capacitance C load and two resistors R inhi and R exci is similar to a damped resistive bridge.
- the potential V load at the propagation terminal Bp is then gradually loaded as shown in the figure 7 .
- This figure 7 illustrates that the integration of the synaptic weight can only be effective over a period of integration PI corresponding to a reading pulse duration less than the time constant of the couple R syn - C load .
- V load V read whatever the value of R exci .
- This time constant is also even shorter when binary RRAMs are used to obtain multivalued synapses, the paralleling of these binary RRAMs leading in effect to a lower overall resistance.
- This problem can be circumvented by using a static reading of the synapse, that is to say reading a quantity depending on the weight of the synapse which is stabilized in the sense that it is independent of the duration. of a read pulse.
- V load in the static state reached with a sufficiently long reading pulse, the value of V load can actually be modulated by the ratio between R exci and R inhi and therefore by playing with the number of devices in high resistance ( “High Resistive State” (HRS) or in low resistance (“Low Resistive State”, LRS) on the excitatory and inhibitory component side.
- HRS High Resistive State
- LRS Low Resistive State
- the grids SL a , SL b are intended to remain polarized at ⁇ V read and the write pulses between the ground and V DD are sent to the activation terminals WL i , WL j . It is noted that the access transistors of the inhibitory cells are reverse biased.
- the exciting component of a synapse is taken between a word line WL i_exci , WL j_exci and a bit line Bl u -Bl x while the inhibiting component of the same synapse is taken between a word line WL i_inhi , WL j_inhi and the bit line Bl u -Bl x .
- Read pulses are sent simultaneously on the word lines WL i_exci and WL i_inhi corresponding to a presynaptic spike pertaining to line i.
- the word line WL i_exci undergoes a pulse of the mass at V read and the word line WL i_inhi undergoes a pulse of the mass at -V read .
- these source lines SL a and SL b can change polarization (from V read to -V read and vice versa) only when necessary by coming to exploit for this purpose a memory in each source line pilot.
- the figures 10 and 11 illustrate another exemplary embodiment for integrating an analog signed weight in a synaptic plane according to which an analog absolute value is associated with a binary sign by means of a synapse called in the following synapse absolute value + sign.
- the synaptic plane is then more classic, equipped with a single read voltage and having unidirectional read currents.
- a certain number of binary RRAMs, a multivalued RRAM or a single binary RRAM in the event of binary synapse implements (s) the absolute value R abs of synaptic weight while another RRAM, modeled by a resistance R sign , is used binary to encode the sign of synaptic weight.
- the artificial neuron reading circuit then includes a binary reading circuit Cb of the synaptic weight sign memory M sign and an analog reading circuit Ca of the synaptic weight absolute value memory M abs , said analog reading circuit Ca being controlled by the binary read circuit Cb.
- the binary reading circuit Cb uses the sign of the presynaptic spike Sp and the sign of the synaptic weight M sign to provide the analog reading circuit Ca with an integration sign Si indicating in which direction to integrate the analog current I abs from the absolute value memory of synaptic weight M abs .
- the analog quantity representative of the synaptic weight supplied by the read circuit to the integration circuit is a voltage.
- the analog magnitude representative of the synaptic weight can be a voltage independent of the duration of a reading pulse, the neurons performing a static reading of the synapse.
- the reading circuit CL1 comprises a synaptic current accumulator C load at the terminals of which a voltage V load is established , this voltage corresponding to the analog quantity representative of the synaptic weight.
- the artificial neuron also comprises a bidirectional voltage follower circuit ST1, ST2 interposed between the read circuit CL1 and the integration circuit CI1.
- the neuron works as follows.
- the lower pole of C load is connected to the virtual ground Gnd by a rst_bot switch.
- C load is first discharged by connecting its upper pole to the virtual ground using an rst_top switch, the read circuit being thus configured to discharge the accumulator from synaptic current C load before imposing the reading voltage on the synapse.
- the reading of the synapse is then triggered by the opening of the rst_top switch.
- the current from the bit line BL is then accumulated in C load , generating the read voltage V load . Opening rst_bot provides a floating V load voltage.
- the voltage follower ST1, ST2 is then configured to copy the membrane voltage V mem to the lower terminal of C load .
- the voltage at the upper limit of C load is worth the addition of V mem and V load .
- the voltage follower ST1, ST2 is configured to copy this addition of voltage to the upper limit of the membrane capacity C mem thus providing the new membrane voltage which is compared to the spike emission threshold (s). (s).
- the analog voltage addition proposed in the exemplary embodiments of Figures 11 and 12 allows a static reading of the 2R1C synapse. Dynamic reading is also possible provided that a sufficiently short reading pulse is generated.
- the maximum stimulation is limited by the reading voltage of the RRAMs. We can then choose that such maximum stimulation alone can trigger a postsynaptic spike by a neuron whose membrane tension is at its resting potential. In which case, the voltage range of the membrane voltage V mem is limited by the read voltage V read . Alternatively, such a stimulation characteristic is not provided, and the membrane tension V mem can be spread over a wider voltage range.
- the voltage follower ST1 includes two operational amplifiers, one intended via the mem2bot switch to copy the membrane voltage to the lower terminal of C load and the other intended via the top2mem switch to copy the voltage addition at the upper bound of C mem .
- the voltage follower ST2 includes a single operational amplifier in place of the two operational amplifiers used alternately on the figure 12 .
- an additional capacity C load_bis is provided in parallel with the switch rst_bot.
- the reading circuit CL2 comprises a switch R ON / R OFF having a first terminal B1 intended to be connected to the synapse via the bit line BL and a second terminal B2 connected to the integration circuit CI2.
- the integration circuit CI2 includes an integrator assembly connected to the second terminal and within which is mounted the synaptic weight accumulator C mem .
- the synapse is read as follows. Initially, the switch is left open (it has a resistance R OFF in its blocked state), the voltage on the bit line V BL stabilizes according to the weight of the synapse (we must therefore have R OFF >> HRS ). In a second step, the switch is closed (it has a resistance R ON in its passing state), the voltage V BL is integrated by the integrator circuit, the time constant of which depends on R ON and C mem and not on the LRS of RRAMs. So we have R ON > LRS. The neuron of the figure 14 thus performs a static reading of the synapse by means of the switch R ON / R OFF and integration of the voltage V BL .
- the neuron comprises a logic circuit LOG interposed between the read circuit CL3 and the integration circuit Cl3, the logic circuit being configured to generate from said analog quantity a pulse Cmd_exci, Cmd_inhi having said duration.
- the integration circuit Cl3 for its part comprises a current source SI exc , SI inhi piloted by said pulse to inject a current into the accumulator of synaptic weights during said duration.
- the reading circuit CL3 of the neuron represented in figure 15 comprises a synaptic current accumulator C load at the terminals of which a voltage V load is established and an rst_load switch making it possible to discharge this accumulator before a reading operation of the synaptic weight.
- the C load accumulator is charged by the synaptic current, more or less quickly depending on the synaptic weight. The duration of this loading can be estimated very simply using an analog comparator.
- the output Comp_exci of the comparator Cp is supplied to the logic circuit LOG interposed between the read circuit CL3 and the integration circuit CI3.
- This logic circuit is configured to generate, from the result of the comparison, a voltage pulse having said duration, in this case the Cmd_exci pulse in the example of the figure 15 .
- the integration circuit CI2 for its part comprises a current source SI exc controlled by said pulse Cmd_exci to inject a current into the accumulator of synaptic weights C mem during said duration.
- the duration of this current injection therefore depends on the value of the synaptic weight, so that the final value of the membrane tension V mem also depends on the value of the synaptic weight.
- the integration circuit CI2 includes another current source SI inhi piloted by a voltage pulse Cmd_inhi to draw current from the synaptic weight accumulator C mem for a period representative of the synaptic weight.
- the neuron of the figure 15 can also take into account the sign of the presynaptic pulse, this Signe_spike being supplied to the logic circuit LOG which is then configured to determine the integration sign from the spike sign and weight sign information synaptic synapse determined by which of the comparators Cp and Cn which switches.
- This logical and economical treatment of the spike sign allows, particularly in the case of 1T1R cells, to lighten the synaptic plane and its management. It will be noted that the synapse absolute value + sign is particularly suitable for this neuron.
- An advantage of this embodiment is the reliability of zero weights.
- all RRAMs devices in the synapse are at HRS. So, on the one hand, the time constant associated with a zero weight synapse is much more important than the others and, on the other hand, we will have Rexci ⁇ Rinhi. It is therefore unlikely that the voltage V load exceeds one of the two thresholds Sp, Sn not only because it will not have time but also because its final value risks being between the two thresholds. Consequently, the logic circuit LOG sees no switching at the output of the comparators Cp, Cn and does not trigger any charge injection on C mem . A true zero weight which does not in any way modify the state of the neuron is thus implemented.
- the synapse is read in a direction long enough for V load to start abutting on V read or -V read according to the sign of the synapse. Then, the polarization of the reading is reversed, the voltage V load will therefore cross the potential difference from one polarization of V read to the other and cross the two thresholds Sp, Sn in a time relative to the weight of the synapse.
- the logic circuit LOG is then configured to generate a voltage pulse between these two comparator switching operations.
- the reading circuit CL4, CL5 comprises a current conveyor which comprises an input port X intended to be connected to the synapse via the bit line, an input port Y intended to be subjected to a fixed voltage (by example ground) for the duration of access to the synapse and an output port Z connected to the integration circuit.
- Such a current conveyor based on an operational amplifier and two current mirrors, is such that the fixed voltage imposed on the input port Y is copied over the input port X and the current injected into the input port X is reproduced on the output port Z (with a sign inversion for the circuit of the figure 17 , without inversion of sign for the circuit of the figure 18 ).
- a fixed voltage being imposed on the Y port we just slave the bit line voltage to this fixed voltage (it is thus independent of the membrane voltage of the neuron) and debit the synaptic current on the output port Z. This synaptic current is supplied to the integration circuit which will then carry out an analog integration.
- the current conveyor comprises, as shown in figure 19 , a current mirror configured to achieve a lowering of the current between the input port X and the output port Z.
- I OUT - I IN K , with K a factor greater than 1 which makes it possible to use reading pulses of longer duration and therefore more suited to the operation of the active elements without thereby saturating the membrane capacity.
- the neuron comprises a transmission gate T between the read circuit and the integration circuit.
- This gate makes it possible to dissociate the access time to the RRAMs and the integration time, which makes it possible to wait for the voltage on the input port X to stabilize before integrating the reading current and to play with a brief integration pulse without requiring a particularly reactive operational amplifier.
- the reading circuit CL6 advantageously comprises both the reversing current mirror of the reading circuit CL4 and the non-reversing current mirror of the reading circuit CL5 (these mirrors being able to effect a lowering of the current in accordance with the figure 19 ) and a selector S controlled by a signal Signe_spike representative of the sign of the presynaptic spike to select the output of one or other of these current mirrors.
- the invention is not limited to the above-described artificial neuron, but extends to a neuromorphic chip comprising a plurality of synapses with resistive memory arranged in a network with transverse lines and columns. Each synapse has a propagation terminal, the propagation terminals of the synapses of the same column being linked together and connected to an artificial neuron according to the invention.
- the synapses can be excitatory or inhibitory synapses such as those presented above in connection with the figures 6 to 11 .
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biomedical Technology (AREA)
- Theoretical Computer Science (AREA)
- Biophysics (AREA)
- Molecular Biology (AREA)
- General Health & Medical Sciences (AREA)
- Neurology (AREA)
- Evolutionary Computation (AREA)
- Data Mining & Analysis (AREA)
- Computational Linguistics (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- Computer Hardware Design (AREA)
- Semiconductor Memories (AREA)
- Heterocyclic Carbon Compounds Containing A Hetero Ring Having Oxygen Or Sulfur (AREA)
Abstract
L'invention porte sur un neurone artificiel pour puce neuromorphique comprenant une synapse à mémoire résistive représentative d'un poids synaptique. Le neurone artificiel comprend un circuit de lecture, un circuit d'intégration et un circuit logique interposé entre le circuit de lecture et le circuit d'intégration.Le circuit de lecture (CL3) est configuré pour imposer à la synapse une tension de lecture indépendante de la tension de membrane et pour fournir une grandeur analogique représentative du poids synaptique.Le circuit logique (LOG) est configuré pour générer à partir de ladite grandeur analogique une impulsion (Cmd_exci, Cmd_inhi) présentant ladite duréeLe circuit d'intégration comprend un accumulateur de poids synaptiques aux bornes duquel s'établit une tension de membrane (V<sub>mem</sub>) et un comparateur configuré pour émettre une impulsion postsynaptique en cas de franchissement d'un seuil par la tension de membrane. Il comprend par ailleurs une source de courant (SI<sub>exc</sub>, SI<sub>inhi</sub>) pilotée par ladite impulsion pour injecter un courant dans l'accumulateur de poids synaptiques pendant ladite durée.The invention relates to an artificial neuron for a neuromorphic chip comprising a synapse with resistive memory representative of a synaptic weight. The artificial neuron comprises a reading circuit, an integration circuit and a logic circuit interposed between the reading circuit and the integration circuit.The reading circuit (CL3) is configured to impose on the synapse an independent reading voltage of the membrane voltage and to provide an analog quantity representative of the synaptic weight. The logic circuit (LOG) is configured to generate from said analog quantity a pulse (Cmd_exci, Cmd_inhi) exhibiting said duration The integration circuit comprises an accumulator of synaptic weights at the terminals of which a membrane voltage is established (V <sub> mem </sub>) and a comparator configured to emit a postsynaptic pulse in the event of a threshold being crossed by the membrane voltage. It also includes a current source (SI <sub> exc </sub>, SI <sub> inhi </sub>) controlled by said pulse to inject current into the synaptic weight accumulator during said duration.
Description
Le domaine de l'invention est celui des puces neuromorphiques à réseaux de neurones artificiels exploitant des synapses à mémoire résistive.The field of the invention is that of neuromorphic chips with artificial neural networks exploiting synapses with resistive memory.
Une cellule nerveuse, ou neurone, peut être décomposée en plusieurs parties :
- Les dendrites qui sont les entrées du neurone via lesquelles il reçoit des signaux d'excitation ou d'inhibition ;
- Le corps du neurone qui est le théâtre d'échanges ioniques à travers la membrane cellulaire ;
- L'axone, une longue extension du corps cellulaire, qui est l'unique sortie du corps.
- The dendrites which are the inputs of the neuron via which it receives signals of excitation or inhibition;
- The body of the neuron which is the scene of ionic exchanges across the cell membrane;
- The axon, a long extension of the cell body, which is the only exit from the body.
Selon les signaux d'excitation ou d'inhibition reçus au niveau des dendrites, des ions transitent à travers la membrane cellulaire. Le déséquilibre de charges entre l'intérieur et l'extérieur de la cellule induit une différence de tension de part et d'autre de la membrane. On parle alors de tension de membrane aux bornes d'une capacité de membrane. Lorsque cette tension de membrane dépasse un certain niveau, c'est-à-dire lorsque la cellule est suffisamment excitée, le neurone connait un échange brutal d'ions. Il en résulte une variation significative de la tension de membrane. Cette variation, appelée potentiel d'action (« action potential » ou encore « spike » en anglais), se propage le long de l'axone, vers les boutons synaptiques qui constituent les sorties du neurone. Vus de l'extérieur de la cellule, ces « spikes » constituent l'activité électrique du neurone.Depending on the excitation or inhibition signals received at the dendrites, ions pass through the cell membrane. The imbalance of charges between the inside and the outside of the cell induces a voltage difference on either side of the membrane. We then speak of membrane voltage across a membrane capacity. When this membrane tension exceeds a certain level, that is to say when the cell is sufficiently excited, the neuron experiences a brutal exchange of ions. This results in a significant variation in the membrane tension. This variation, called action potential (“action potential” or “spike” in English), propagates along the axon, towards the synaptic buttons which constitute the outputs of the neuron. Seen from outside the cell, these "spikes" constitute the electrical activity of the neuron.
Dans un réseau de neurones biologiques, chaque neurone est connecté à plusieurs milliers d'autres via autant de synapses. Le terme synapse désigne la connexion entre la terminaison axonique d'un neurone dit présynaptique et une dendrite d'un neurone dit postsynaptique. L'influence du neurone présynaptique sur le neurone postsynaptique est pondérée par le poids de la synapse qui peut être excitatrice ou inhibitrice. Dans le premier cas, un « spike » présynaptique charge la tension de membrane du neurone postsynaptique et précipite la génération d'un potentiel d'action postsynaptique. Dans le second cas, un « spike » présynaptique a pour effet de dépolariser la membrane postsynaptique et de retarder l'apparition d'un potentiel d'action postsynaptique.In a network of biological neurons, each neuron is connected to several thousand others via as many synapses. The term synapse designates the connection between the axon termination of a so-called presynaptic neuron and a dendrite of a so-called postsynaptic neuron. The influence of the presynaptic neuron on the postsynaptic neuron is balanced by the weight of the synapse which can be excitatory or inhibitory. In the first case, a presynaptic spike charges the membrane tension of the postsynaptic neuron and precipitates the generation of a postsynaptic action potential. In the second case, a Presynaptic "spike" has the effect of depolarizing the postsynaptic membrane and delaying the onset of a postsynaptic action potential.
Les réseaux de neurones artificiels sont utilisés dans différents domaines de traitement du signal comme par exemple en classification de données, en reconnaissance d'images ou en prise de décision. Ils sont inspirés des réseaux de neurones biologiques dont ils imitent le fonctionnement et sont essentiellement composés de neurones artificiels interconnectés entre eux par des synapses qui peuvent être implémentées par des composants résistifs dont la conductance varie en fonction de la tension appliquée à leurs bornes.Artificial neural networks are used in different fields of signal processing, such as data classification, image recognition or decision-making. They are inspired by the biological neural networks of which they imitate the functioning and are essentially composed of artificial neurons interconnected by synapses which can be implemented by resistive components whose conductance varies according to the voltage applied to their terminals.
De par leur grande densité et leur caractère non-volatil, ces composants résistifs (memristors ou RRAM) sont des candidats idéaux à l'implémentation des synapses. La résistance variable de ces composants peut être augmentée (opération dite Reset) ou diminuée (opération Set) si on y applique des grandeurs électriques (tension et/ou courant) relativement élevées. Si on veut simplement lire la valeur de leur résistance sans la modifier (opération Read), il faut appliquer des grandeurs électriques relativement faibles.Due to their high density and their non-volatile nature, these resistive components (memristors or RRAM) are ideal candidates for the implementation of synapses. The variable resistance of these components can be increased (so-called Reset operation) or decreased ( Set operation) if relatively high electrical quantities (voltage and / or current) are applied. If you simply want to read the value of their resistance without modifying it ( Read operation), you must apply relatively small electrical quantities.
L'intégration de synapses résistives prend souvent la forme d'un plan mémoire, qu'on nomme « plan synaptique », dans lequel les synapses sont agencées en un réseau à lignes et colonnes transversales qui permet de relier une couche de neurones d'entrée (les neurones présynaptiques) et une couche de neurones de sortie (les neurones postsynaptiques). Chaque synapse dispose d'une borne d'activation de la synapse et d'une borne de propagation d'un signal synaptique. Les bornes d'activation des synapses d'une même ligne sont reliées entre elles par l'intermédiaire d'une ligne de mot (ou « Word-Line »), et les bornes de propagation des synapses d'une même colonne sont reliées entre elles et connectées à un neurone artificiel par l'intermédiaire d'une ligne de bit (ou « Bit-Line »). Une ligne de mot est utilisée pour injecter une impulsion en tension dans les synapses de la ligne correspondante et les lignes de bit sont les sorties de ces synapses. En présence d'une activation présynaptique sur une ligne de mot depuis un neurone présynaptique, chaque ligne de bit propage un courant pondéré par la valeur de la mémoire résistive correspondante vers un neurone postynaptique.The integration of resistive synapses often takes the form of a memory plane, which is called "synaptic plane", in which the synapses are arranged in a network with transverse lines and columns which makes it possible to connect a layer of input neurons. (presynaptic neurons) and a layer of output neurons (postsynaptic neurons). Each synapse has a synapse activation terminal and a propagation terminal for a synaptic signal. The terminals for activating synapses on the same line are linked together via a word line (or “Word-Line”), and the terminals for propagating synapses in the same column are connected between and connected to an artificial neuron via a bit line (or "Bit-Line"). A word line is used to inject a voltage pulse into the synapses of the corresponding line and the bit lines are the outputs of these synapses. In the presence of a presynaptic activation on a word line from a presynaptic neuron, each bit line propagates a current weighted by the value of the corresponding resistive memory towards a postynaptic neuron.
Comme représenté en
Comme représenté en
Un certain nombre de contraintes pèsent sur l'implémentation effective d'un tel réseau de neurones artificiels.A certain number of constraints weigh on the effective implementation of such an artificial neural network.
Par exemple, si l'organisation du plan synaptique présente des avantages (parallélisation des calculs, densité, mise en commun des pilotes), elle présente néanmoins des limitations. Tout d'abord, cette organisation offre une connexion « tous-à-tous» selon laquelle tous les neurones présynaptiques sont connectés via une synapse à tous les neurones postsynaptiques. Or dans certains cas, cette connexion « tous-à-tous » peut conduire à une matrice synaptique principalement nulle et donc à une faible efficacité énergétique. Cette organisation nécessite également de devoir traiter finement les poids nuls sous peine d'avoir un réseau composé principalement de synapses parasites, rendant encore plus complexe l'intégration analogique des poids synaptiques.For example, if the organization of the synaptic plan has advantages (parallel calculations, density, pooling of pilots), it nevertheless has limitations. First of all, this organization offers an “all-to-all” connection according to which all the presynaptic neurons are connected via a synapse to all the postsynaptic neurons. However in certain cases, this “all-to-all” connection can lead to a mainly zero synaptic matrix and therefore to low energy efficiency. This organization also requires having to process zero weights at the risk of having a network composed mainly of parasitic synapses, making the analog integration of synaptic weights even more complex.
Une autre contrainte est liée à l'implémentation classique d'un neurone artificiel sous la forme d'un neurone intégratif à seuil (« Integrate and Fire » en anglais). Le principe d'une telle implémentation est représenté sur la
Or selon la loi d'Ohm, et comme représenté en
Ceci implique que plus la tension Vmem est élevée, donc plus le neurone est excité et susceptible d'émettre un spike, moins ses synapses afférentes ont un effet sur lui. Un tel phénomène est problématique. Ainsi, dans une couche de classification par exemple, le neurone censé émettre le plus de spikes voit son activité réduite dans une plus grande proportion que celle des neurones moins actifs. Dès lors, la différence entre ce neurone « gagnant » et les autres neurones se trouve réduite.This implies that the higher the voltage V mem , the more the neuron is excited and likely to emit a spike, the less its afferent synapses have an effect on it. Such a phenomenon is problematic. Thus, in a classification layer for example, the neuron supposed to emit the most spikes sees its activity reduced in a greater proportion than that of the less active neurons. Consequently, the difference between this “winning” neuron and the other neurons is reduced.
D'autres contraintes sont liées aux caractéristiques des mémoires RRAM. Un exemple de telle mémoire RRAM est la mémoire OxRAM qui est composée d'un isolant coincé entre deux couches de conducteurs. Lors d'une première étape de formation, un fil conducteur croît depuis l'un des conducteurs vers l'autre, réduisant ainsi la résistance totale du dipôle. Une fois ce fil formé, les actions d'écriture agissent sur sa longueur pour faire varier la résistance. Une opération Set allonge le fil et réduit la résistance tandis qu'une opération Reset a l'effet inverse. La lecture d'une mémoire OxRAM présente une difficulté, celle de l'écriture parasite en venant petit à petit modifier la valeur de la résistance au fil de lectures successives. Pour limiter ce risque, il est conseillé d'appliquer une tension de lecture maximale de 100 mV. La petitesse de cette tension complique grandement la mise en œuvre d'une lecture analogique intégrative de ces mémoires. En effet, la tension de lecture Vspike_pre limite la tension de membrane Vmem, car la capacité de membrane Cmem est chargée à travers la résistance Rsynb sur laquelle est appliquée Vspike_pre. Autrement dit, la tension de membrane vaut au plus la tension Vspike_pre. Or Vmem étant un élément de calcul analogique donc soumis notamment à des problématiques de bruit, de variations technologiques, de couplage capacitif, sa gamme de tension ne peut être réduite à une valeur aussi faible que 100 mV sans augmenter fortement les sources d'erreurs de calcul.Other constraints are linked to the characteristics of RRAM memories. An example of such an RRAM memory is the OxRAM memory which is composed of an insulator wedged between two layers of conductors. During a first stage of formation, a conductive wire grows from one of the conductors towards the other, thus reducing the total resistance of the dipole. Once this wire is formed, the writing actions act on its length to vary the resistance. A Set operation lengthens the wire and reduces the resistance, while a Reset operation has the opposite effect. Reading an OxRAM memory presents a difficulty, that of parasitic writing by gradually changing the value of the resistance over successive readings. To limit this risk, it is advisable to apply a maximum reading voltage of 100 mV. The smallness of this voltage greatly complicates the implementation of an integrative analog reading of these memories. Indeed, the reading voltage V spike_pre limits the membrane voltage V mem , because the membrane capacity C mem is charged through the resistor R synb on which V spike_pre is applied. In other words, the membrane tension is worth at most the tension V spike_pre . However, V mem being an analog calculation element therefore subject in particular to problems of noise, technological variations, capacitive coupling, its voltage range cannot be reduced to a value as low as 100 mV without greatly increasing the sources of errors. Calculation.
On relèvera par ailleurs que les dispositifs RRAMs ont souvent une gamme de résistance faible (avec des valeurs basses de l'ordre du kΩ) et que l'on peut donc être confronté à des constantes de temps du coupe Rsyn - Cmem de l'ordre de la nanoseconde voire moins. Or il n'est pas aisé de générer des impulsions de lecture inférieures à 1 ns, sauf à avoir recours à un circuit spécifique. L'impulsion de lecture doit en outre être propagée sur toute la largeur du plan synaptique. Or plus cette impulsion est courte, moins elle est « propre » en bout de ligne. Enfin, une telle impulsion se propage difficilement dans des circuits analogiques actifs chargés d'intégrer le courant de lecture sans être confrontée à des problématiques de bande passante et/ou de distorsion.It will also be noted that RRAMs devices often have a low resistance range (with low values of the order of kΩ) and that we can therefore be confronted with time constants of the cut R syn - C mem of l nanosecond order or less. However, it is not easy to generate reading pulses of less than 1 ns, except when using a specific circuit. The reading impulse must also be propagated over the entire width of the synaptic plane. The shorter this impulse, the less “clean” it is in the end. Finally, such a pulse is difficult to propagate in active analog circuits responsible for integrating the read current without being confronted with problems of bandwidth and / or distortion.
L'invention a pour objectif d'apporter une solution permettant de répondre à l'une et/ou l'autre des contraintes susmentionnées. Elle propose pour ce faire un neurone artificiel pour puce neuromorphique comprenant une synapse à mémoire résistive représentative d'un poids synaptique, le neurone artificiel comportant un circuit d'intégration qui comprend un accumulateur de poids synaptiques aux bornes duquel s'établit une tension de membrane et un comparateur configuré pour émettre une impulsion postsynaptique en cas de franchissement d'un seuil par la tension de membrane.The invention aims to provide a solution to meet one and / or the other of the above constraints. To do this, it proposes an artificial neuron for a neuromorphic chip comprising a synapse with resistive memory representative of a synaptic weight, the artificial neuron comprising an integration circuit which comprises an accumulator of synaptic weights at the terminals of which a membrane tension is established. and a comparator configured to emit a postsynaptic pulse in the event of a threshold being crossed by the membrane voltage.
Le neurone comporte en outre :
- un circuit de lecture configuré pour imposer à la synapse une tension de lecture indépendante de la tension de membrane et pour fournir une grandeur analogique représentative du poids synaptique, ladite grandeur analogique étant une durée ; et
- un circuit logique interposé entre le circuit de lecture et le circuit d'intégration, le circuit logique étant configuré pour générer à partir de ladite grandeur analogique une impulsion présentant ladite durée.
- a reading circuit configured to impose on the synapse a reading voltage independent of the membrane voltage and to provide an analog quantity representative of the synaptic weight, said analog quantity being a duration; and
- a logic circuit interposed between the read circuit and the integration circuit, the logic circuit being configured to generate from said analog quantity a pulse having said duration.
Le circuit d'intégration comprend une source de courant pilotée par ladite impulsion pour injecter un courant dans l'accumulateur de poids synaptiques pendant ladite durée. Certains aspects préférés mais non limitatifs de ce neurone artificiel sont les suivants :
- le circuit de lecture comprend un accumulateur de courant synaptique, au moins un comparateur de la tension aux bornes de l'accumulateur de courant synaptique avec un seuil pour fournir un résultat de comparaison et le circuit logique est configuré pour générer ladite impulsion à partir du résultat de la comparaison ;
- l'au moins un comparateur du circuit de lecture comprend un premier comparateur à un premier seuil et un deuxième comparateur à un deuxième seuil et le circuit logique est configuré pour déterminer un signe d'intégration en exploitant un signal représentatif d'un signe d'impulsion présynaptique et une information de signe de poids synaptique déterminée par celui des premier et deuxième comparateurs qui commute ;
- l'au moins un comparateur du circuit de lecture comprend un premier comparateur à un premier seuil et un deuxième comparateur à un deuxième seuil et le circuit logique est configuré de telle manière que la durée de ladite impulsion corresponde à une durée séparant le franchissement de l'un des seuils puis le franchissement de l'autre des seuils ;
- la synapse comprenant une mémoire de valeur absolue de poids synaptique et une mémoire de signe de poids synaptique, le circuit de lecture comprend un circuit de lecture binaire de la mémoire de signe de poids synaptique et un circuit de lecture analogique de la mémoire de valeur absolue de poids synaptique piloté par le circuit de lecture binaire.
- the reading circuit comprises a synaptic current accumulator, at least one comparator of the voltage across the terminals of the synaptic current accumulator with a threshold to provide a comparison result and the logic circuit is configured to generate said pulse from the result of the comparison;
- the at least one comparator of the reading circuit comprises a first comparator at a first threshold and a second comparator at a second threshold and the logic circuit is configured to determine an integration sign by exploiting a signal representative of a sign of presynaptic pulse and synaptic weight sign information determined by which of the first and second comparators switches;
- the at least one comparator of the reading circuit comprises a first comparator at a first threshold and a second comparator at a second threshold and the logic circuit is configured in such a way that the duration of said pulse corresponds to a duration separating the crossing of the 'one of the thresholds then crossing the other of the thresholds;
- the synapse comprising a synaptic weight absolute value memory and a synaptic weight sign memory, the reading circuit comprises a binary reading circuit of the synaptic weight sign memory and an analog reading circuit of the absolute value memory of synaptic weight controlled by the binary reading circuit.
D'autres aspects, buts, avantages et caractéristiques de l'invention apparaîtront mieux à la lecture de la description détaillée suivante de formes de réalisation préférées de celle-ci, donnée à titre d'exemple non limitatif, et faite en référence aux dessins suivants sur lesquels, outre les
- la
figure 5 est un schéma de principe d'un neurone artificiel selon l'invention ; - la
figure 6 illustre une synapse excitatrice ou inhibitrice pouvant être utilisée dans le cadre de l'invention ; - la
figure 7 représente la réponse impulsionnelle de la synapse de lafigure 6 ; - les
figures 8 et9 fournissent des exemples d'intégration de la synapse de lafigure 6 dans un plan synaptique ; - les
figures 10 et11 fournissent un autre exemple de synapse excitatrice ou inhibitrice pouvant être utilisée dans le cadre de l'invention ; - les
figures 12 et13 sont des schémas de neurones artificiels disposant d'un accumulateur de courant synaptique dans le circuit de lecture et d'un accumulateur de poids synaptiques dans le circuit d'intégration et pour lesquels la grandeur analogique représentative du poids synaptique fournie par le circuit de lecture au circuit d'intégration est une tension ; - la
figure 14 représente un autre exemple de neurone pour lequel la grandeur analogique est également une tension ; - la
figure 15 représente un neurone selon l'invention pour lequel la grandeur analogique est également une durée ; - la
figure 16 illustre un autre mode opératoire du neurone de lafigure 15 ; - les
figures 17, 18 et20 représentent des circuits de lecture d'un exemple de neurone pour lequel la grandeur analogique est un courant ; - la
figure 19 illustre une variante de réalisation des circuits de lectures desfigures 17 et 18 .
- the
figure 5 is a block diagram of an artificial neuron according to the invention; - the
figure 6 illustrates an excitatory or inhibitory synapse which can be used within the framework of the invention; - the
figure 7 represents the impulse response of the synapse of thefigure 6 ; - the
figures 8 and9 provide examples of integration of the synapse of thefigure 6 in a synaptic plane; - the
figures 10 and11 provide another example of an excitatory or inhibitory synapse which can be used in the context of the invention; - the
figures 12 and13 are diagrams of artificial neurons with a synaptic current accumulator in the reading circuit and an accumulator synaptic weights in the integration circuit and for which the analog quantity representative of the synaptic weight supplied by the read circuit to the integration circuit is a voltage; - the
figure 14 represents another example of a neuron for which the analog quantity is also a voltage; - the
figure 15 represents a neuron according to the invention for which the analog quantity is also a duration; - the
figure 16 illustrates another operating mode of the neuron of thefigure 15 ; - the
figures 17, 18 and20 represent read circuits of an example of a neuron for which the analog quantity is a current; - the
figure 19 illustrates an alternative embodiment of the reading circuits of theFigures 17 and 18 .
En référence à la
La synapse dispose d'une borne d'activation de la synapse et d'une borne de propagation d'un signal synaptique. En présence d'une activation présynaptique appliquée sur la borne d'activation, la borne de propagation propage un signal synaptique représentatif de la valeur de la mémoire résistive (i.e. le poids synaptique) en direction du neurone artificiel NA par l'intermédiaire d'une ligne de bit BL.The synapse has a terminal for activating the synapse and a terminal for propagating a synaptic signal. In the presence of a presynaptic activation applied to the activation terminal, the propagation terminal propagates a synaptic signal representative of the value of the resistive memory (ie the synaptic weight) towards the artificial neuron NA via a BL bit line.
Le neurone artificiel comprend un circuit d'intégration CI qui comprend un accumulateur de poids synaptiques ACC aux bornes duquel s'établit une tension de membrane Vmem et un comparateur COMP configuré pour émettre une impulsion postsynaptique SO en cas de franchissement d'un seuil Vseuil par la tension de membrane Vmem.The artificial neuron includes an integration circuit CI which includes an accumulator of synaptic weights ACC at the terminals of which a membrane voltage V mem is established and a comparator COMP configured to emit a postsynaptic pulse SO in the event of crossing a threshold V threshold by the membrane voltage V mem .
L'invention propose de découpler la lecture du poids synaptique de son intégration en évitant de numériser la lecture du poids synaptique. Les différentes contraintes peuvent ainsi être réparties selon qu'elles sont héritées de la technologie de la mémoire résistive ou du fonctionnement du neurone analogique. Pour ce faire, l'invention propose plus particulièrement que le neurone artificiel NA comporte en outre un circuit de lecture CL configuré pour imposer à la synapse une tension de lecture indépendante de la tension de membrane Vmem et pour fournir au circuit d'intégration CI une grandeur analogique représentative du poids synaptique PSa.The invention proposes to decouple the reading of the synaptic weight from its integration while avoiding digitizing the reading of the synaptic weight. The different constraints can thus be distributed according to whether they are inherited from resistive memory technology or from the functioning of the analog neuron. To do this, the invention proposes more particularly that the artificial neuron NA furthermore comprises a reading circuit CL configured to impose on the synapse a reading voltage independent of the membrane voltage V mem and to supply to the integration circuit CI an analog quantity representative of the synaptic weight PSa.
Le circuit de lecture CL permet ainsi d'imposer à la synapse une tension de lecture présentant une polarisation identique pour chaque lecture quel que soit l'état de la tension de membrane Vmem. Ce circuit de lecture permet également d'imposer à la synapse une tension de lecture suffisamment faible pour éviter tout risque d'écriture parasite. Le résultat de la lecture, à savoir la grandeur analogique représentative du poids synaptique PSa, est transmis au circuit d'intégration. Comme on le verra par la suite, cette grandeur peut prendre différentes formes, telles qu'une tension, un courant, une charge ou une durée par exemple. Le circuit d'intégration Cl, ainsi découplé de la lecture, peut ainsi proposer une vaste plage de tension pour la tension de membrane, indépendante des spécifications de la synapse.The reading circuit CL thus makes it possible to impose on the synapse a reading voltage having an identical polarization for each reading regardless of the state of the membrane voltage V mem . This reading circuit also makes it possible to impose a sufficiently low reading voltage on the synapse to avoid any risk of parasitic writing. The result of the reading, namely the analog quantity representative of the synaptic weight PSa, is transmitted to the integration circuit. As will be seen below, this quantity can take different forms, such as a voltage, a current, a charge or a duration for example. The integration circuit Cl, thus decoupled from the reading, can thus propose a wide range of voltage for the membrane voltage, independent of the specifications of the synapse.
On détaille dans ce qui suit des implémentations de synapses résistives qui trouvent avantageusement application dans le cadre de l'invention. On comprendra cependant que l'invention n'est pas limitée à ces implémentations spécifiques.We detail in the following implementations of resistive synapses which advantageously find application in the context of the invention. It will however be understood that the invention is not limited to these specific implementations.
On peut en effet chercher à disposer de synapses qui puissent être à la fois excitatrices et inhibitrices, et qui doivent donc être capable à la fois d'injecter et de tirer du courant depuis le neurone. Les synapses étant ici résistives et le courant qu'elles pilotent dépendant de leur résistance, on prévoit deux niveaux de tension d'alimentation de la synapse, un positif et un négatif. La
Le poids de la synapse est encodé comme suit. Si on veut une synapse excitatrice, la résistance Rinhi est réglée à son niveau de résistance maximal, tandis que la résistance Rexci (<Rinhi) module le poids positif de la synapse. Plus la résistance Rexci est faible, plus le poids synaptique est important, car il en résulte un courant synaptique supérieur. A contrario, si on veut une synapse inhibitrice, la résistance Rexci est réglée à son niveau de résistance maximal, tandis que la résistance Rinhi (<Rexci) module le poids négatif de la synapse
En négligeant les effets des sélecteurs ou transistors d'accès, le schéma composé de la capacité Cload et de deux résistances Rinhi et Rexci s'apparente à un pont résistif amorti. En appliquant les tensions d'alimentation Vread et -Vread, le potentiel Vload à la borne de propagation Bp se charge alors progressivement comme représenté sur la
By neglecting the effects of the access selectors or transistors, the diagram composed of the capacitance C load and two resistors R inhi and R exci is similar to a damped resistive bridge. By applying the supply voltages V read and -V read , the potential V load at the propagation terminal Bp is then gradually loaded as shown in the
Cette constante de temps est en outre encore plus brève lorsque l'on a recours à des RRAM binaires pour obtenir des synapses multivaluées, la mise en parallèle de ces RRAM binaires débouchant en effet sur une résistance globale plus faible. Ce problème peut être contourné en ayant recours à une lecture statique de la synapse, c'est-à-dire à la lecture d'une grandeur fonction du poids de la synapse qui est stabilisée en ce sens qu'elle est indépendante de la durée d'une impulsion de lecture. La
Prenant l'exemple d'une synapse composée de 8 dispositifs OxRAM, 4 pour la composante excitatrice et 4 pour la composante inhibitrice. Avec LRS = 3kΩ, HRS = 800kΩ et Cload = 50 fF, un codage sur 9 poids (dont le poids nul) associés à des constantes de temps allant de 1 ns à 250 ps peut être obtenu. Une lecture statique offre quant à elle plus de combinaisons (13 poids possible) et évite d'avoir à générer des impulsions de lectures brèves. En revanche, cette lecture statique est plus énergivore en raison de l'accès « prolongé » à la RRAM.Taking the example of a synapse composed of 8 OxRAM devices, 4 for the excitatory component and 4 for the inhibitory component. With LRS = 3kΩ, HRS = 800kΩ and C load = 50 fF, coding on 9 weights (including zero weight) associated with time constants ranging from 1 ns to 250 ps can be obtained. A static reading offers more combinations (13 possible weights) and avoids having to generate short reading pulses. On the other hand, this static reading is more energy-consuming because of the “prolonged” access to the RRAM.
On a représenté sur les
Dans le cas 1T1R (
Dans le cas 1S1R (
On peut également rechercher à implémenter des spikes négatifs, à savoir des impulsions présynaptiques qui inversent le poids de la synapse qu'elles déclenchent.One can also seek to implement negative spikes, namely presynaptic impulses which reverse the weight of the synapse which they trigger.
Reprenant l'exemple de la
Reprenant l'exemple de la
Les
On a représenté sur les
Sur les
Le fonctionnement du neurone est le suivant. Lors d'une opération de lecture de la synapse, le pôle inférieur de Cload est connecté à la masse virtuelle Gnd par un interrupteur rst_bot. Dans une première phase de l'opération de lecture, Cload est d'abord déchargée en reliant son pôle supérieur à la masse virtuelle à l'aide d'un interrupteur rst_top, le circuit de lecture étant ainsi configuré pour décharger l'accumulateur de courant synaptique Cload avant d'imposer la tension de lecture à la synapse. La lecture de la synapse est ensuite déclenchée par l'ouverture de l'interrupteur rst_top. Le courant provenant de la ligne de bit BL est alors accumulé dans Cload, générant la tension de lecture Vload. L'ouverture de rst_bot permet d'obtenir une tension Vload flottante. Le suiveur de tension ST1, ST2 est alors configuré pour recopier la tension de membrane Vmem à la borne inférieure de Cload. Ainsi la tension à la borne supérieure de Cload vaut l'addition de Vmem et Vload. Puis le suiveur de tension ST1, ST2 est configuré pour recopier cette addition de tension à la borne supérieure de la capacité de membrane Cmem fournissant ainsi la nouvelle tension de membrane qui est comparée au(x) seuil(s) d'émission de spike(s).The neuron works as follows. During a synapse read operation, the lower pole of C load is connected to the virtual ground Gnd by a rst_bot switch. In a first phase of the read operation, C load is first discharged by connecting its upper pole to the virtual ground using an rst_top switch, the read circuit being thus configured to discharge the accumulator from synaptic current C load before imposing the reading voltage on the synapse. The reading of the synapse is then triggered by the opening of the rst_top switch. The current from the bit line BL is then accumulated in C load , generating the read voltage V load . Opening rst_bot provides a floating V load voltage. The voltage follower ST1, ST2 is then configured to copy the membrane voltage V mem to the lower terminal of C load . Thus the voltage at the upper limit of C load is worth the addition of V mem and V load . Then the voltage follower ST1, ST2 is configured to copy this addition of voltage to the upper limit of the membrane capacity C mem thus providing the new membrane voltage which is compared to the spike emission threshold (s). (s).
L'addition analogique de tension proposée dans les exemples de réalisation de
Sur la
Sur la
On relèvera que la synapse valeur absolue + signe est adaptée aux neurones des
Sur la
La lecture de la synapse s'opère comme suit. Dans un premier temps, on laisse le commutateur ouvert (il présente une résistance ROFF dans son état bloqué), la tension sur la ligne de bit VBL se stabilise suivant le poids de la synapse (on doit donc avoir ROFF>>HRS). Dans un second temps, le commutateur est fermé (il présente une résistance RON dans son état passant), la tension VBL est intégrée par le montage intégrateur dont la constante de temps dépend de RON et Cmem et pas du LRS des RRAMs. On a donc RON>LRS. Le neurone de la
On a représenté sur les
Comme pour les neurones des
La sortie Comp_exci du comparateur Cp est fourni au circuit logique LOG interposé entre le circuit de lecture CL3 et le circuit d'intégration CI3. Ce circuit logique est configuré pour générer, à partir du résultat de la comparaison, une impulsion en tension présentant ladite durée, en l'occurrence l'impulsion Cmd_exci dans l'exemple de la
Le circuit d'intégration CI2 comprend quant à lui une source de courant SIexc pilotée par ladite impulsion Cmd_exci pour injecter un courant dans l'accumulateur de poids synaptiques Cmem pendant ladite durée. La durée de cette injection de courant dépend donc de la valeur du poids synaptique, de sorte que la valeur finale de la tension de membrane Vmem dépend également de la valeur du poids synaptique.The integration circuit CI2 for its part comprises a current source SI exc controlled by said pulse Cmd_exci to inject a current into the accumulator of synaptic weights C mem during said duration. The duration of this current injection therefore depends on the value of the synaptic weight, so that the final value of the membrane tension V mem also depends on the value of the synaptic weight.
Dans l'exemple représenté à synapses excitatrices et inhibitrices, le circuit d'intégration CI2 comprend une autre source de courant SIinhi pilotée par une impulsion en tension Cmd_inhi pour tirer du courant depuis l'accumulateur de poids synaptiques Cmem pendant une durée représentative du poids synaptique.In the example shown with excitatory and inhibitory synapses, the integration circuit CI2 includes another current source SI inhi piloted by a voltage pulse Cmd_inhi to draw current from the synaptic weight accumulator C mem for a period representative of the synaptic weight.
Le neurone de la
Un avantage de cette réalisation est la fiabilité des poids nuls. Dans le cas d'un poids nul, tous les dispositifs RRAMs de la synapse sont à HRS. Donc, d'une part, la constante de temps associée à une synapse de poids nulle est bien plus importante que les autres et, d'une autre part, on aura Rexci ≈ Rinhi. Il est donc peu probable que la tension Vload franchise l'un des deux seuils Sp, Sn non seulement parce qu'elle n'en aura pas le temps mais aussi parce que sa valeur finale risque d'être comprise entre les deux seuils. Dès lors le circuit logique LOG ne voit pas de commutation en sortie des comparateurs Cp, Cn et ne déclenche aucune injection de charge sur Cmem. Un véritable poids nul qui ne modifie absolument en rien l'état du neurone est ainsi implémenté.An advantage of this embodiment is the reliability of zero weights. In the case of zero weight, all RRAMs devices in the synapse are at HRS. So, on the one hand, the time constant associated with a zero weight synapse is much more important than the others and, on the other hand, we will have Rexci ≈ Rinhi. It is therefore unlikely that the voltage V load exceeds one of the two thresholds Sp, Sn not only because it will not have time but also because its final value risks being between the two thresholds. Consequently, the logic circuit LOG sees no switching at the output of the comparators Cp, Cn and does not trigger any charge injection on C mem . A true zero weight which does not in any way modify the state of the neuron is thus implemented.
On a représenté sur la
Ainsi, dans un premier temps, la synapse est lue dans un sens suffisamment longtemps pour que Vload parte en buttée sur Vread ou -Vread selon le signe de la synapse. Puis, on inverse la polarisation de la lecture, la tension Vload va donc traverser la différence de potentiel d'une polarisation de Vread à l'autre et franchir les deux seuils Sp, Sn en un temps relatif au poids de la synapse. Le circuit logique LOG est alors configuré pour générer une impulsion en tension entre ces deux commutations de comparateurs.Thus, at first, the synapse is read in a direction long enough for V load to start abutting on V read or -V read according to the sign of the synapse. Then, the polarization of the reading is reversed, the voltage V load will therefore cross the potential difference from one polarization of V read to the other and cross the two thresholds Sp, Sn in a time relative to the weight of the synapse. The logic circuit LOG is then configured to generate a voltage pulse between these two comparator switching operations.
On a représenté sur les
Un tel convoyeur de courant, basé sur un amplificateur opérationnel et deux miroirs de courant, est tel que la tension fixe imposée sur le port d'entrée Y est recopiée sur le port d'entrée X et le courant injecté dans le port d'entrée X est reproduit sur le port de sortie Z (avec une inversion de signe pour le circuit de la
La constante de temps Rsyn-Cmem peut toutefois s'avérer trop brève pour que l'amplificateur opérationnel et les miroirs de courant composant le convoyeur de courant fonctionnement correctement. Ainsi, dans une variante de réalisation, le convoyeur de courant comprend, comme représenté en
Dans une variante de réalisation représentée sur la
Par ailleurs, sur cette
L'invention n'est pas limitée au neurone artificiel précédemment décrit, mais s'étend à une puce neuromorphique comprenant une pluralité de synapses à mémoire résistive agencées en un réseau à lignes et colonnes transversales. Chaque synapse dispose d'une borne de propagation, les bornes de propagation des synapses d'une même colonne étant reliées entre elles et connectées à un neurone artificiel selon l'invention. Les synapses peuvent être des synapses excitatrices ou inhibitrices telles que celles présentées précédemment en liaison avec les
Claims (7)
le neurone artificiel comportant un circuit d'intégration (CI3) qui comprend un accumulateur de poids synaptiques (Cmem) aux bornes duquel s'établit une tension de membrane (Vmem) et un comparateur (COMP) configuré pour émettre une impulsion postsynaptique en cas de franchissement d'un seuil par la tension de membrane,
le neurone artificiel étant caractérisé en ce qu'il comporte en outre :
the artificial neuron comprising an integration circuit (CI3) which comprises an accumulator of synaptic weights (C mem ) at the terminals of which a membrane tension is established (V mem ) and a comparator (COMP) configured to emit a postsynaptic pulse in membrane threshold crossing,
the artificial neuron being characterized in that it further comprises:
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
FR1872515A FR3089663B1 (en) | 2018-12-07 | 2018-12-07 | Artificial neuron for neuromorphic chip with resistive synapses |
Publications (2)
Publication Number | Publication Date |
---|---|
EP3663988A1 true EP3663988A1 (en) | 2020-06-10 |
EP3663988B1 EP3663988B1 (en) | 2022-05-18 |
Family
ID=66049326
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
EP19212046.7A Active EP3663988B1 (en) | 2018-12-07 | 2019-11-28 | Artificial neuron for neuromorphic chip with resistive synapses |
Country Status (3)
Country | Link |
---|---|
US (1) | US11630993B2 (en) |
EP (1) | EP3663988B1 (en) |
FR (1) | FR3089663B1 (en) |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2021262023A1 (en) * | 2020-06-25 | 2021-12-30 | PolyN Technology Limited | Analog hardware realization of neural networks |
RU2796649C2 (en) * | 2020-06-25 | 2023-05-29 | ПолиН Текнолоджи Лимитед | Analogue hardware implementation of neural networks |
US11823037B1 (en) | 2023-02-28 | 2023-11-21 | PolyN Technology Limited | Optocoupler-based flexible weights in neuromorphic analog signal processors |
US11885271B2 (en) | 2020-06-25 | 2024-01-30 | PolyN Technology Limited | Systems and methods for detonation control in spark ignition engines using analog neuromorphic computing hardware |
Families Citing this family (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US11188815B2 (en) | 2019-01-07 | 2021-11-30 | International Business Machines Corporation | Weight shifting for neuromorphic synapse array |
KR20210056476A (en) * | 2019-11-08 | 2021-05-20 | 삼성전자주식회사 | Synapse element and neuromorphic processor including synapse element |
KR102507770B1 (en) * | 2020-08-26 | 2023-03-07 | 국민대학교산학협력단 | Synaptic device and driving method thereof, and computing system using the same and driving method thereof |
WO2022118340A1 (en) * | 2020-12-03 | 2022-06-09 | Indian Institute Of Technology Delhi | Novel activation function with hardware realization for recurrent neuromorphic networks |
US11741353B2 (en) | 2020-12-09 | 2023-08-29 | International Business Machines Corporation | Bias scheme for single-device synaptic element |
CN113408613B (en) * | 2021-06-18 | 2022-07-19 | 电子科技大学 | Single-layer image classification method based on delay mechanism |
CN113643741B (en) * | 2021-08-16 | 2023-12-15 | 湖北大学 | 1S 1R-based logic operation unit and operation method |
KR102579571B1 (en) * | 2021-12-03 | 2023-09-15 | 아주대학교산학협력단 | Pixelated monolithic photoplethysmography sensor with spiking neural network structure and method for operating the same |
WO2023128792A1 (en) * | 2021-12-30 | 2023-07-06 | PolyN Technology Limited | Transformations, optimizations, and interfaces for analog hardware realization of neural networks |
CN115392442B (en) * | 2022-10-26 | 2023-01-24 | 深圳时识科技有限公司 | Neuron ignition method, chip and electronic equipment |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20170243108A1 (en) * | 2016-02-19 | 2017-08-24 | International Business Machines Corporation | Current Mirror Scheme for An Integrating Neuron Circuit |
US20180260696A1 (en) * | 2017-03-08 | 2018-09-13 | Arm Ltd | Spiking neural network |
WO2018201060A1 (en) * | 2017-04-27 | 2018-11-01 | The Regents Of The University Of California | Mixed signal neuromorphic computing with nonvolatile memory devices |
Family Cites Families (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20120011092A1 (en) * | 2010-07-07 | 2012-01-12 | Qualcomm Incorporated | Methods and systems for memristor-based neuron circuits |
US11501141B2 (en) * | 2018-10-12 | 2022-11-15 | Western Digital Technologies, Inc. | Shifting architecture for data reuse in a neural network |
-
2018
- 2018-12-07 FR FR1872515A patent/FR3089663B1/en active Active
-
2019
- 2019-11-28 EP EP19212046.7A patent/EP3663988B1/en active Active
- 2019-12-04 US US16/703,465 patent/US11630993B2/en active Active
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20170243108A1 (en) * | 2016-02-19 | 2017-08-24 | International Business Machines Corporation | Current Mirror Scheme for An Integrating Neuron Circuit |
US20180260696A1 (en) * | 2017-03-08 | 2018-09-13 | Arm Ltd | Spiking neural network |
WO2018201060A1 (en) * | 2017-04-27 | 2018-11-01 | The Regents Of The University Of California | Mixed signal neuromorphic computing with nonvolatile memory devices |
Non-Patent Citations (3)
Title |
---|
EERO LEHTONEN ET AL: "Memristive Synapses Are Becoming Reality Master's thesis View project Gyrocardiography View project", 31 January 2010 (2010-01-31), XP055614483, Retrieved from the Internet <URL:https://www.researchgate.net/profile/Piotr_Dudek4/publication/49288092_Memristive_Synapses_Are_Becoming_Reality/links/55759d8b08aeb6d8c01978c2/Memristive-Synapses-Are-Becoming-Reality.pdf> [retrieved on 20190821] * |
OLGA KRESTINSKAYA ET AL: "Neuro-memristive Circuits for Edge Computing: A review", 28 November 2018 (2018-11-28), XP055614484, Retrieved from the Internet <URL:https://arxiv.org/pdf/1807.00962.pdf> [retrieved on 20190821], DOI: 10.1109/TNNLS.2019.2899262 * |
VAN SCHAIK A ED - ROY ASIM ET AL: "Building blocks for electronic spiking neural networks", NEURAL NETWO, ELSEVIER SCIENCE PUBLISHERS, BARKING, GB, vol. 14, no. 6-7, 9 July 2001 (2001-07-09), pages 617 - 628, XP004310067, ISSN: 0893-6080, DOI: 10.1016/S0893-6080(01)00067-3 * |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2021262023A1 (en) * | 2020-06-25 | 2021-12-30 | PolyN Technology Limited | Analog hardware realization of neural networks |
RU2796649C2 (en) * | 2020-06-25 | 2023-05-29 | ПолиН Текнолоджи Лимитед | Analogue hardware implementation of neural networks |
US11885271B2 (en) | 2020-06-25 | 2024-01-30 | PolyN Technology Limited | Systems and methods for detonation control in spark ignition engines using analog neuromorphic computing hardware |
US11823037B1 (en) | 2023-02-28 | 2023-11-21 | PolyN Technology Limited | Optocoupler-based flexible weights in neuromorphic analog signal processors |
Also Published As
Publication number | Publication date |
---|---|
FR3089663B1 (en) | 2021-09-17 |
EP3663988B1 (en) | 2022-05-18 |
FR3089663A1 (en) | 2020-06-12 |
US20200202206A1 (en) | 2020-06-25 |
US11630993B2 (en) | 2023-04-18 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
EP3663988B1 (en) | Artificial neuron for neuromorphic chip with resistive synapses | |
EP3660749A1 (en) | Neural circuit suitable for implementing synaptic learning | |
EP2965269B1 (en) | Artificial neuron and memristor | |
EP2819068B1 (en) | Artificial neuron including a resistive memory | |
EP2713318B1 (en) | Neuromorphic system exploiting the intrinsic characteristics of memory cells | |
WO2013000940A1 (en) | Network of artificial neurones based on complementary memristive devices | |
WO2013000939A1 (en) | Method for non-supervised learning in an artificial neurone network based on memristive nanodevices, and artificial neurone network implementing said method | |
FR2970589A1 (en) | VOLATILE MEMORY CELL / NON VOLATILE | |
EP3154061B1 (en) | Method and circuit for controlling the programming current in a non-volatile memory array | |
EP2689422A1 (en) | Logical memory architecture, in particular for mram, pcram, or rram | |
EP4224477A1 (en) | Three-dimensional memory structure for in-memory computing | |
EP3549070B1 (en) | Modulation device and method, artificial synapse comprising said modulation device, short-term plasticity method in an artificial neural network comprising said artificial synapse | |
EP4020479A1 (en) | Differential reading of rram memory with low power consumption | |
EP4137999A1 (en) | Neuromorphic circuit made of 2t2r rram cells | |
EP4292017A1 (en) | Neuromorphic circuit and associated training method | |
FR3130447A1 (en) | OxRAM/FeRAM mixed technology non-volatile memories | |
FR3024272A1 (en) | NON-VOLATILE MEMORY WITH PROGRAMMABLE RESISTANCE | |
EP2987167B1 (en) | Non-volatile memory cell | |
EP4354351A1 (en) | Electronic circuit based on rram cells | |
EP4283621A1 (en) | Method for calculating a mac operation in a 1s1r rram | |
FR3135562A1 (en) | Memory cell, electronic circuit comprising such cells, programming method and associated multiplication and accumulation method | |
EP4345821A1 (en) | Logic data processing circuit integrated in a data storage circuit | |
FR3125163A1 (en) | METHOD OF READING A MULTI-LEVEL RRAM | |
FR2970591A1 (en) | VOLATILE MEMORY CELL AND NON-VOLATILE COMBINED | |
FR3035998A1 (en) | NON-VOLATILE MEMORY WITH PROGRAMMING CIRCUIT |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PUAI | Public reference made under article 153(3) epc to a published international application that has entered the european phase |
Free format text: ORIGINAL CODE: 0009012 |
|
STAA | Information on the status of an ep patent application or granted ep patent |
Free format text: STATUS: REQUEST FOR EXAMINATION WAS MADE |
|
17P | Request for examination filed |
Effective date: 20191128 |
|
AK | Designated contracting states |
Kind code of ref document: A1 Designated state(s): AL AT BE BG CH CY CZ DE DK EE ES FI FR GB GR HR HU IE IS IT LI LT LU LV MC MK MT NL NO PL PT RO RS SE SI SK SM TR |
|
AX | Request for extension of the european patent |
Extension state: BA ME |
|
GRAP | Despatch of communication of intention to grant a patent |
Free format text: ORIGINAL CODE: EPIDOSNIGR1 |
|
STAA | Information on the status of an ep patent application or granted ep patent |
Free format text: STATUS: GRANT OF PATENT IS INTENDED |
|
INTG | Intention to grant announced |
Effective date: 20211221 |
|
GRAS | Grant fee paid |
Free format text: ORIGINAL CODE: EPIDOSNIGR3 |
|
GRAA | (expected) grant |
Free format text: ORIGINAL CODE: 0009210 |
|
STAA | Information on the status of an ep patent application or granted ep patent |
Free format text: STATUS: THE PATENT HAS BEEN GRANTED |
|
AK | Designated contracting states |
Kind code of ref document: B1 Designated state(s): AL AT BE BG CH CY CZ DE DK EE ES FI FR GB GR HR HU IE IS IT LI LT LU LV MC MK MT NL NO PL PT RO RS SE SI SK SM TR |
|
REG | Reference to a national code |
Ref country code: GB Ref legal event code: FG4D Free format text: NOT ENGLISH |
|
REG | Reference to a national code |
Ref country code: CH Ref legal event code: EP |
|
REG | Reference to a national code |
Ref country code: IE Ref legal event code: FG4D Free format text: LANGUAGE OF EP DOCUMENT: FRENCH |
|
REG | Reference to a national code |
Ref country code: DE Ref legal event code: R096 Ref document number: 602019015069 Country of ref document: DE |
|
REG | Reference to a national code |
Ref country code: AT Ref legal event code: REF Ref document number: 1493586 Country of ref document: AT Kind code of ref document: T Effective date: 20220615 |
|
REG | Reference to a national code |
Ref country code: LT Ref legal event code: MG9D |
|
REG | Reference to a national code |
Ref country code: NL Ref legal event code: MP Effective date: 20220518 |
|
REG | Reference to a national code |
Ref country code: AT Ref legal event code: MK05 Ref document number: 1493586 Country of ref document: AT Kind code of ref document: T Effective date: 20220518 |
|
PG25 | Lapsed in a contracting state [announced via postgrant information from national office to epo] |
Ref country code: SE Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20220518 Ref country code: PT Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20220919 Ref country code: NO Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20220818 Ref country code: NL Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20220518 Ref country code: LT Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20220518 Ref country code: HR Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20220518 Ref country code: GR Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20220819 Ref country code: FI Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20220518 Ref country code: ES Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20220518 Ref country code: BG Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20220818 Ref country code: AT Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20220518 |
|
PG25 | Lapsed in a contracting state [announced via postgrant information from national office to epo] |
Ref country code: RS Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20220518 Ref country code: PL Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20220518 Ref country code: LV Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20220518 Ref country code: IS Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20220918 |
|
PG25 | Lapsed in a contracting state [announced via postgrant information from national office to epo] |
Ref country code: SM Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20220518 Ref country code: SK Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20220518 Ref country code: RO Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20220518 Ref country code: EE Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20220518 Ref country code: DK Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20220518 Ref country code: CZ Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20220518 |
|
REG | Reference to a national code |
Ref country code: DE Ref legal event code: R097 Ref document number: 602019015069 Country of ref document: DE |
|
PLBE | No opposition filed within time limit |
Free format text: ORIGINAL CODE: 0009261 |
|
STAA | Information on the status of an ep patent application or granted ep patent |
Free format text: STATUS: NO OPPOSITION FILED WITHIN TIME LIMIT |
|
PG25 | Lapsed in a contracting state [announced via postgrant information from national office to epo] |
Ref country code: AL Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20220518 |
|
26N | No opposition filed |
Effective date: 20230221 |
|
PG25 | Lapsed in a contracting state [announced via postgrant information from national office to epo] |
Ref country code: SI Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20220518 |
|
PG25 | Lapsed in a contracting state [announced via postgrant information from national office to epo] |
Ref country code: MC Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20220518 |
|
REG | Reference to a national code |
Ref country code: CH Ref legal event code: PL |
|
REG | Reference to a national code |
Ref country code: BE Ref legal event code: MM Effective date: 20221130 |
|
PG25 | Lapsed in a contracting state [announced via postgrant information from national office to epo] |
Ref country code: LI Free format text: LAPSE BECAUSE OF NON-PAYMENT OF DUE FEES Effective date: 20221130 Ref country code: CH Free format text: LAPSE BECAUSE OF NON-PAYMENT OF DUE FEES Effective date: 20221130 |
|
PG25 | Lapsed in a contracting state [announced via postgrant information from national office to epo] |
Ref country code: LU Free format text: LAPSE BECAUSE OF NON-PAYMENT OF DUE FEES Effective date: 20221128 |
|
PG25 | Lapsed in a contracting state [announced via postgrant information from national office to epo] |
Ref country code: IE Free format text: LAPSE BECAUSE OF NON-PAYMENT OF DUE FEES Effective date: 20221128 |
|
PG25 | Lapsed in a contracting state [announced via postgrant information from national office to epo] |
Ref country code: BE Free format text: LAPSE BECAUSE OF NON-PAYMENT OF DUE FEES Effective date: 20221130 |
|
PG25 | Lapsed in a contracting state [announced via postgrant information from national office to epo] |
Ref country code: IT Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20220518 |
|
PGFP | Annual fee paid to national office [announced via postgrant information from national office to epo] |
Ref country code: FR Payment date: 20231121 Year of fee payment: 5 Ref country code: DE Payment date: 20231120 Year of fee payment: 5 |
|
PG25 | Lapsed in a contracting state [announced via postgrant information from national office to epo] |
Ref country code: HU Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT; INVALID AB INITIO Effective date: 20191128 |
|
PG25 | Lapsed in a contracting state [announced via postgrant information from national office to epo] |
Ref country code: CY Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20220511 |
|
PG25 | Lapsed in a contracting state [announced via postgrant information from national office to epo] |
Ref country code: MK Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20220511 |